Natural gas provides more than one-fifth of all the primary energy used in the United States. Much raw gas is "subquality", that is, it exceeds the pipeline specifications in nitrogen, carbon dioxide and/or hydrogen sulfide content. A representative range of U.S. gas compositions, compared to the specifications that must be met to bring the gas to pipeline quality, is shown in Table 1.
TABLE 1 ______________________________________ Natural Gas Compositions Found in the U.S. and the Specifications of the Pipeline Grid Component Typical pipeline spec. Typical composition range ______________________________________ Hydrogen &lt;4 ppm ** 76% &lt;4 ppm sulfide 11% 4-1,000 ppm 4% 1,000-10,000 ppm 8% &gt;10,000 ppm Carbon &lt;1-3% ** 72% &lt;1% dioxide 18% 1-3% 7% 3-10% 3% &gt;10% Water 80-140 ppm 800-1,200 ppm Inerts (CO.sub.2, &lt;3-4% -- N.sub.2, He, Ar, etc.) Oxygen &lt;0.4% -- ______________________________________ ** Compositions for producing wells. Unexploited reserves contain higher fractions of subquality gas.
The best treatment for natural gas right now is no treatment. Currently, more than half of the gas produced in the U.S. can be brought to pipeline specification after minimal processing, such as glycol dehydration to remove water. Raw gas that is known to be high in nitrogen content, high in nitrogen plus carbon dioxide content, or high in hydrogen sulfide content is usually left in the ground, because it cannot be extracted and treated economically with present processing technology.
There are several aspects to the problem of treating natural gas to bring it to pipeline specifications. The first is the removal of impurities, primarily water, hydrogen sulfide and carbon dioxide; the second is loss of methane during processing. Processes that remove hydrogen sulfide and carbon dioxide may also remove a portion of the methane. Losses of less than about 3% are normally acceptable; losses of 3-10% may be acceptable if offset by other advantages; losses above 10% are normally unacceptable. A third aspect is the fate of the impurities once removed. Carbon dioxide can be discharged or reinjected, but hydrogen sulfide, which is toxic even in low concentrations, must be treated. If the waste stream containing hydrogen sulfide can be concentrated sufficiently, it may be passed to a Claus plant for conversion to sulfur. Waste streams containing low concentrations must be disposed of in some other way, such as a redox process of the LO CAT or Stretford type, for example, or, less desirably, flaring.
Choice of appropriate treatment is, therefore, not straightforward, and depends on the feed gas composition, the size and location of the plant and other variables.
When natural gas is treated, most plants handling large volumes of sour gas containing greater than about 200 ppm hydrogen sulfide use amine-based technology for acid gas removal. Amines commonly used include MEA, DEA, DIPA, DGA and MDEA. The plants can remove both carbon dioxide and hydrogen sulfide. When the amine solution is spent, the acid gases are flashed off and the solution is regenerated. The mechanical equipment in an amine plant makes it susceptible to failure. The plant includes heaters, aerial coolers, pumps, etc. and requires frequent quality checks and maintenance, making operational reliability probably the weakest feature of the technology.
Amine plants do not sorb methane to any significant extent, so methane loss is not an issue in this case. However, the hydrogen-sulfide-containing gas stream produced when the sorbent is regenerated must still be treated, subject to the same constraints as above.
As an alternative to amine sorption, or as a polishing step following any process, specialized scavenging or sulfur recovery processes, such as Sulfa-Scrub, Sulfa-Check, Chemsweet, Supertron 600, solid iron sponge or solid zinc oxide may be used for low-volume streams containing less than about 100 ppm hydrogen sulfide. Many scavengers present substantial disposal problems, however. In an increasing number of states, the spent scavenger constitutes toxic waste.
A considerable body of literature exists regarding membrane-based treatment of natural gas, mostly using cellulose acetate (CA) membranes to remove carbon dioxide. Membrane technology is attractive for this separation, because treatment can be accomplished using the high wellhead gas pressure as the driving force for the separation. Membrane systems, have, however, been slow to penetrate the market, and it is estimated that no more than about 1% of all processing is carried out using membranes. Nevertheless, for small-scale streams of appropriate composition, cellulose acetate membrane plants are state-of-the art, and up to 100 of these are believed to have been installed. Although all of these plants are designed to remove carbon dioxide, cellulose acetate membranes also have selectivity for hydrogen sulfide over methane, so they tend to coextract small amounts of hydrogen sulfide. Unless the raw gas stream contains very high concentrations of carbon dioxide, however, it is not possible to reduce a stream containing even modest amounts of hydrogen sulfide to pipeline specification (usually 4 ppm hydrogen sulfide) without vastly overprocessing as far as the carbon dioxide specification is concerned. If such overprocessing is performed, large amounts of methane are lost in the membrane permeate stream, and this is normally unacceptable.
Only a few of the many literature references relating to membrane-based carbon dioxide treatment specifically discuss removal of hydrogen sulfide in conjunction with the carbon dioxide. A paper by W. J. Schell et al. ("Separation of CO.sub.2 from Mixtures by Membrane Permeation", presented at the Gas Conditioning Conference, University of Oklahoma, March 1983) says that "If the H.sub.2 S level is low enough, the membrane system can also be used to meet pipeline specification for this component without any further treatment required." The paper shows a case where a cellulose acetate membrane system can be used to reach pipeline specification for carbon dioxide and hydrogen sulfide in two stages, starting with a feed content of 15% carbon dioxide and 250 ppm hydrogen sulfide, and points out that, for high concentrations of hydrogen sulfide, "a much larger number of elements are required to reduce the H.sub.2 S levels to pipeline specification (1/4 grain) than for CO.sub.2 (3%)." The costs of membrane treatment are estimated to be more than 100% higher than conventional amine treatment in this case.
A report by N. N. Li et al. to the Department of Energy ("Membrane Separation Processes in the Petrochemical Industry", Phase II Final Report, September 1987) examined the effect of impurities, including hydrogen sulfide, on the ability of cellulose acetate membranes to remove carbon dioxide from natural gas. The reporters found that the membrane performance was not affected significantly by hydrogen sulfide alone. However, dramatic loss of membrane permeability was observed if both hydrogen sulfide and water vapor were present in the feed. The authors concluded that "successful use of these CA-based membranes must avoid processing gas which simultaneously has high H.sub.2 O and H.sub.2 S concentrations".
Another problem associated with cellulose acetate membranes is water, which is always present in raw natural gas streams to some extent, as vapor, entrained liquid, or both. The gas separation properties of cellulose acetate membranes are destroyed by contact with liquid water, so it is normally necessary to provide pretreatment to knock out any liquid water and to reduce the relative humidity low enough that there is no risk of condensation of water within the membrane modules on the permeate side. For example, the above-cited paper by W. J. Schell et al. ("Separation of CO.sub.2 from Mixtures by Membrane Permeation", presented at the Gas Conditioning Conference, University of Oklahoma, March 1983) points out that "Even though membrane systems simultaneously dehydrate while removing CO.sub.2, care must be taken to avoid contacting the membrane with liquid water. Feed gas streams saturated with water are normally preheated to at least 10.degree. above the water dew point at the feed inlet pressure and the pressure tubes and inlet piping are insulated to prevent condensation."
The above-cited report by N. N. Li et al. ("Membrane Separation Processes in the Petrochemical Industry, Phase II Final Report, September 1987) presents data showing the effect of water vapor on membrane flux for cellulose acetate membranes, and concludes that "for relative humidities of 30% and higher, the flux decline is large, rapid, and irreversible". E. W. Funk et al. ("Effect of Impurities on Cellulose Acetate Membrane Performance", Recent Advances in Separation Techniques-III, AIChE Symposium Series, 250, Vol 82, 1986) advocate that "Moisture levels up to 20% RH appear tolerable but higher levels can cause irreversible membrane compaction".
U.S. Pat. No. 4,130,403 to T. E. Cooley et al. (Removal of H.sub.2 S and/or CO.sub.2 from a Light Hydrocarbon Stream by Use of Gas Permeable Membrane, 1978, Col. 12, lines 36-39) states that "It has been discovered that in order to function effectively, the feed gas to the cellulose ester membrane should be substantially water free". A second paper by W. J. Schell et al. ("Spiral-Wound Permeators for Purification and Recovery", Chemical Engineering Progress, October 1982, pages 33-37) confirms that "Liquid water is detrimental to the performance of the membrane, however, so that the feed gas is delivered to the membrane system at less than 90% relative humidity."
In other words, although cellulose acetate membranes will permeate water preferentially over methane, and hence have the capability to dehydrate the gas stream, care must be taken to keep the amounts of water vapor being processed low, and, according to some teachings, as low as 20-30% relative humidity.
In light of these limitations, considerable effort has been expended over the last few years in the search for membrane materials that would be better able to handle streams containing carbon dioxide plus secondary contaminants, notably hydrogen sulfide and water. Developers of membrane materials usually rely on the solution-diffusion model to predict vapor and gas permeation behavior. In this model, it is assumed that gas at the high-pressure side of the membrane dissolves in the membrane material and diffuses down a concentration gradient to the low-pressure side of the membrane, where the gas is desorbed. It is assumed that the gas phases on either side of the membrane are in thermodynamic equilibrium with their respective polymeric interfaces, and that the interfacial sorption and desorption processes are rapid compared to the rate of diffusion through the membrane. Thus, the rate-limiting step is diffusion. For simple gases, Fick's law leads to the equation: ##EQU1## which can be further simplified to ##EQU2## where J is the membrane flux (cm.sup.3 (STP)/cm.sup.2.s.cmHg), D is the diffusion coefficient of the gas in the membrane (cm.sup.2 /sec) and is a measure of the gas mobility, L is the membrane thickness, k is the Henry's law sorption coefficient linking the concentration of the gas in the membrane material to the pressure in the adjacent gas (cm.sup.3 (STP)/cm.sup.3.cmHg), and .DELTA.p is the pressure difference across the membrane (cmHg).
The gas permeability ##EQU3## is a property of the membrane material and is independent of membrane thickness. In this definition, permeability is a measure of the rate at which a particular gas moves through a membrane of standard thickness (1 cm) under a standard pressure difference (1 cmHg). The convenient permeability unit, 1.times.10.sup.-10 cm.sup.3 (STP).cm/cm.sup.2.s. cmHg, is often called a Barter, after R. M. Barrer, a pioneer in membrane permeation studies.
A measure of the ability of a membrane to separate two gases, A and B, is the ratio of their permeabilities, .alpha., called the membrane selectivity, ##EQU4## Equation (4) can also be written as ##EQU5##
The ratio D.sub.A /D.sub.B is the ratio of the diffusion coefficients of the two gases and can be viewed as the mobility selectivity, reflecting the different sizes of the two molecules. The ratio k.sub.A /k.sub.B is the ratio of the Henry's law solubility coefficients of the two gases and can be viewed as the solubility selectivity, reflecting the relative condensabilities of the two gases.
In all polymer materials, the diffusion coefficient decreases with increasing molecular size, because large molecules interact with more segments of the polymer chain than small molecules. Hence, the mobility coefficient always favors the passage of small molecules over large ones. The sorption coefficient, on the other hand, is a measure of the energy required for the permeant to be sorbed by the polymer and increases with the condensability of the permeant. This dependence on condensability means that the sorption coefficient increases with molecular diameter, because large molecules are normally more condensable than smaller ones. The combined effect of these two factors determines the selectivity of the membrane.
The balance between mobility selectivity and sorption selectivity is different for glassy and rubbery polymers. In glassy polymers, the mobility term is usually dominant, permeability falls with increasing permeant size and small molecules permeate preferentially. In rubbery polymers, the sorption term is usually dominant, permeability increases with increasing permeant size and larger molecules permeate preferentially. Since both carbon dioxide (3.3 .ANG.) and hydrogen sulfide (3.6 .ANG.) have smaller kinetic diameters than methane (3.8 .ANG.), and since both carbon dioxide and hydrogen sulfide are more condensable than methane, both glassy and rubbery membranes are selective for the acid gas components over methane. To date, however, most membrane development work in this area has focused on glassy materials, of which cellulose acetate is the most successful example.
In citing selectivity, it is important to be clear as to how the permeation data being used have been measured. It is common to measure the fluxes of different gases separately, then to calculate selectivity as the ratio of the pure gas permeabilities. This gives the "ideal" selectivity for that pair of gases. Pure gas measurements are more commonly reported than mixed gas experiments, because pure gas experiments are much easier to perform. Measuring the permeation data using gas mixtures, then calculating the selectivity as the ratio of the gas fluxes, gives the actual selectivity that can be achieved under real conditions. In gas mixtures that contain condensable components, it is frequently, although not always, the case that the mixed gas selectivity is lower, and at times considerably lower, than the ideal selectivity. The condensable component, which is readily sorbed into the polymer matrix, swells or, in the case of a glassy polymer, plasticizes the membrane, thereby reducing its discriminating capabilities.
A technique for predicting mixed gas performance under real conditions from pure gas measurements with any reliability has not yet been developed. In the case of gas mixtures such as carbon dioxide/methane with other components, the expectation is that the carbon dioxide at least will have a swelling or plasticizing effect, thereby changing the membrane permeation characteristics. This expectation is borne out by cellulose acetate membranes. For example, according to a paper by M. D. Donahue et al. ("Permeation behavior of carbon dioxide-methane mixtures in cellulose acetate membranes", Journal of Membrane Science, 42, 197-214 1989) when measured with pure gases, the carbon dioxide permeability of asymmetric cellulose acetate is 9.8.times.10.sup.-5 cm.sup.3 /cm.sup.2.s.kPa and the methane permeability is 2.0.times.10.sup.-6 cm.sup.3 /cm.sup.2.s.kPa, giving an ideal selectivity of about 50. Yet the actual selectivity obtained with mixed gases is typically in the range 10-20, a factor of 3-5 times lower than the ideal selectivity. For example, the report to DOE by Norman Li et al., discussed above, gives carbon dioxide/methane selectivities in the range 9-15 for one set of field trials (at 870-905 psi feed pressure) and 12 for another set (at 200 psig feed pressure) with a highly acid feed gas. The W. J. Schell et al. Chemical Engineering Progress paper, discussed above, gives carbon dioxide/methane selectivities of 21 (at 250-450 psig feed pressure) and 23 (at 800 psig feed pressure). Thus, even in mixed gas measurements, a wide spread of selectivities is obtained, the spread depending partly on operating conditions. In particular, the plasticizing or swelling effect of the carbon dioxide on the membrane tends to show pressure dependence, although it is sometimes hard to distinguish this from other effects, such as the contribution of secondary condensable components.
As a first step in developing a new membrane, it is normal to start by testing with pure gases under mild operating conditions. Membranes that appear to pass this test can then be developed and tested further, modified, optimized and scaled up, leading eventually to field tests, full-scale demonstration and ultimately, industrial acceptance. Candidate membranes can and do fall by the wayside at each step along this path, for diverse reasons.
The search for improved membranes for removing acid components from gas streams, although it has focused primarily on glassy membranes, encompasses several types of membranes and membrane materials. A paper by A. Deschamps et al. ("Development of Gaseous Permeation Membranes adapted to the Purification of Hydrocarbons", I.I.F-I.I.R--Commission A3, Paris, 1989) describes work with aromatic polyimides having an intrinsic material selectivity of 80 for carbon dioxide over methane and 200,000 for water vapor over methane. The paper defines the target selectivities that the researchers were aiming for as 50 for carbon dioxide/methane and 200 for water vapor/methane. The paper, which is principally directed to dehydration, does not give carbon dioxide/methane selectivities, except to say that they were "generally low", even though the experiments were carried out with pure gas samples. In other words, despite the high intrinsic selectivity of 80, the lower target value of 50 could not be reached.
British patent number 1,478,083, to Klass and Landahl, presents a large body of permeation data obtained with methane/carbon dioxide/hydrogen sulfide mixed gas streams and polyamide (nylon 6 and nylon 6/6), polyvinyl alcohol (PVA), polyacrylonitrile (PAN) and gelatin membranes. Some unexpectedly high selectivities are shown. For the nylon membranes, carbon dioxide/methane selectivities of up to 30, and hydrogen sulfide/methane selectivities up to 60, are reported. The best carbon dioxide/methane selectivity is 160, for PAN at a temperature of 30.degree. C. and a feed pressure of 65 psia; the best hydrogen sulfide/methane selectivity is 200, for gelatin at the same conditions. In both cases, however, the permeability is extremely low: for carbon dioxide through PAN, less than 5.times.10.sup.-4 Barrer and for hydrogen sulfide through gelatin, less than 3.times.10.sup.-3 Barrer. These low permeabilities would make the transmembrane fluxes miserable for any practical purposes. It is also unknown whether the gelatin membrane, which was plasticized with glycerin, would be stable much above the modest pressures under which it was tested.
U.S. Pat. No. 4,561,864, also to Klass and Landahl, incorporates in its text some of the data reported in the British patent discussed above. The '864 patent also includes a table of calculations for cellulose acetate membranes, showing the relationship between "Figure of Merit", a quantity used to express the purity and methane recovery in the residue stream, as a function of "Flow Rate Factor", a quantity that appears to be somewhat akin to stage-cut. In performing the calculations, separation factors (where the separation factor is the sum of the carbon dioxide/methane selectivity and the hydrogen sulfide/methane selectivity) of 20 to 120 are assumed. The figures used in the calculations appear to range from the low end of the combined carbon dioxide and hydrogen sulfide selectivities from mixed gas data to the high end of the combined selectivities calculated from pure gas data.
A paper by D. L. Ellig et al. ("Concentration of Methane from Mixtures with Carbon Dioxide by permeation through Polymeric Films", Journal of Membrane Science, 6, 259-263, 1980) summarizes permeation tests carried out with 12 different commercially available films and membranes, using a mixed gas feed containing 60% carbon dioxide, 40% methane, but no hydrogen sulfide or water vapor. The tests were carried out at 2,068 kPa (about 300 psi) feed pressure. The results show selectivities of about 9-27 for cellulose acetate, up to 40 for polyethersulfone and 20-30 for polysulfone. One of the membranes tested was nylon, which, in contradiction to the results reported by Klass and Landahl, showed essentially no selectivity at all for carbon dioxide over methane.
The already much-discussed DOE Final Report by N. N. Li et al. contains a section in which separation of polar gases from non-polar gases by means of a mixed-matrix, facilitated transport membrane is discussed. The membrane consists of a silicone rubber matrix carrying polyethylene glycol, which is used to facilitate transport of polar gases, such as hydrogen sulfide, over non-polar gases, such as methane. In tests on natural gas streams, the membranes exhibited hydrogen sulfide/methane selectivity of 25-30 and carbon dioxide/methane selectivity of 7-8, which latter number was considered too low for practical carbon dioxide separation. The membrane was also shown to be physically unstable at feed pressure above about 170 psig, which, even if the carbon dioxide/methane selectivity were adequate, would render it unsuitable for handling raw natural gas streams. U.S. Pat. Nos. 4,608,060, to S. Kulprathipanja, and 4,606,740, to S. Kulprathipanja and S. S. Kulkarni, of Li's group at UOP, present additional data using the same type of glycol-laden membranes as discussed in the DOE report. In this case, however, pure gas tests were performed and ideal hydrogen sulfide/methane selectivities as high as 115-185 are quoted. It is interesting to note that these are 4-8 times higher than the later measured mixed gas numbers quoted in the DOE report. The same effect obtains for carbon dioxide, where the pure gas selectivities are in the range 21-32 and the mixed gas data give selectivities of 7-8.
Similar in concept is U.S. Pat. No. 4,737,166, to S. L. Matson et al., which discloses an immobilized liquid membrane typically containing n-methylpyrrolidone or another polar solvent in cellulose acetate or any other compatible polymer. The membranes and processes discussed in this patent are directed to selective hydrogen sulfide removal, in other words leaving both the methane and the carbon dioxide behind in the residue stream. As in the UOP patents, very high hydrogen sulfide/methane selectivities, in the range 90-350, are quoted. Only pure gas data are given, however, and the feed pressure is 100 psig. There is no discussion as to how the membranes might behave when exposed to multicomponent gas streams and/or high feed pressures. Based on the UOP teachings, the mixed gas, high-pressure results might be expected to be not so good.
U.S. Pat. No. 4,781,733, to C. Babcock et al., describes results obtained with an interfacial composite membrane made by a polycondensation reaction between a diacidchloride-terminated silicone rubber and a diamine. In pure gas experiments at 100 psig, the membrane exhibited hydrogen sulfide/methane selectivities up to 47 and carbon dioxide/methane selectivities up to 50. No mixed gas or high-pressure data are given.
U.S. Pat. No. 4,493,716, to R. H. Swick, reports permeation results obtained with a composite membrane consisting of a polysulfide polymer on a Gorerex (polytetrafluoroethylene) support. Only pure gas, low-pressure test cell permeability data are given. Based on the reported permeabilities, which only give an upper limit for the methane permeability, the membrane appears to have a hydrogen sulfide/methane selectivity of at least 19-42 and a carbon dioxide/methane selectivity of at least 2-6. Some results show that the carbon dioxide permeability increased after exposure to hydrogen sulfide, which might suggest an overall decrease in selectivity if the membrane has become generally more permeable, although no methane data that could confirm or refute this are cited.
U.S. Pat. No. 4,963,165, to I. Blume and I. Pinnau reports pure gas, low-pressure data for a composite membrane consisting of a polyamide-polyether block copolymer on a polyamide support. Hydrogen sulfide/methane selectivities in the range 140-190, and carbon dioxide/methane selectivities in the range 18-20, are quoted. Mixed gas data for a stream containing oxygen, nitrogen, carbon dioxide and sulfur dioxide are also quoted and discussed in the text, but it is not dear how these data would compare with those for methane- or hydrogen-sulfide-containing mixed gas streams.
Despite the many and varied research and development efforts that this body of literature represents, cellulose acetate membranes, with their attendant advantages and disadvantages, remain the only membrane type whose properties in handling acid gas streams under real gas-field operating conditions are reasonably well understood, and the only membrane type in commercial use for removing acid gas components from methane.
U.S. Pat. No. 4,589,896, to M. Chen et al., exemplifies the type of process that must be adopted to remove carbon dioxide and hydrogen sulfide from methane and other hydrocarbons when working within the performance limitations of cellulose acetate membranes. The process is directed at natural gas streams with a high acid gas content, or at streams from enhanced oil recovery (EOR) operations, and consists of a multistage membrane separation, followed by fractionation of the acid gas components and multistage flashing to recover the hydrogen sulfide. The acid-gas-depleted residue stream is also subjected to further treatment to recover hydrocarbons. The raw gas to be treated typically contains as much as 80% or more carbon dioxide, with hydrogen sulfide at the relatively low, few thousands of ppm level. Despite the fact that the ratio of the carbon dioxide content to the hydrogen sulfide content is high (about 400:1), the raw gas stream must be passed through a minimum of four membrane stages, arranged in a three-step, two-stage configuration, to achieve good hydrogen sulfide removal. The goal is not to bring the raw gas stream to natural gas pipeline specification, but rather to recover relatively pure carbon dioxide, free from hydrogen sulfide, for further use in EOR. The target concentration of carbon dioxide in the treated hydrocarbon stream is less than 10%, which would, of course, not meet natural gas pipeline standards. The methane left in the residue stream after higher hydrocarbon removal is simply used to strip carbon dioxide from hydrogen-sulfide-rich solvent in a later part of the separation process; no methane passes to a natural gas pipeline. Despite the multistep/multistage membrane arrangement, in a representative example, about 7% carbon dioxide is left in the hydrocarbon residue stream after processing, and about 12% hydrocarbon loss into the permeate takes place.
A number of references deal with dehydration, but without simultaneously addressing hydrogen sulfide removal. For example, a recent report from W. R. Grace to the Gas Research Institute ("Gas Dehydration by Membranes--Industrial feasibility", T. E. Sulpizio and B. S. Minhas, Project #5729, June 1993) discusses the ability of conventional cellulose acetate and other unspecified experimental membranes to dehydrate natural gas. No data or discussion of streams containing hydrogen sulfide are given. U.S. Pat. Nos. 4,783,201, to Rice and Murphy and 4,497,640 to F. Fournie et al. both describe dehydration of air or methane using various types of membranes. Again, no data or discussion of hydrogen sulfide are included. U.S. Pat. No. 4,781,733, to W. C. Babcock et al., shows water vapor/methane gas mixture data, but reports only pure gas, low-pressure data for hydrogen sulfide.
In summary, it may be seen that there remains a need for improved membranes and improved processes for handling streams containing methane, acid gas components and water vapor.